Hacker Newsnew | past | comments | ask | show | jobs | submit | jaster's commentslogin

All the "Thanks for watching!" gave me a good chuckle.

Remind me of one of my own experiences with one of the Whisper model, where some random noise in the middle of the conversation was translated into "Don't forget to like and subscribe".

Really illustrate where the training data is coming from.


Here is a relevant tidbit from the RWKV paper "Limitations" section (https://arxiv.org/abs/2305.13048):

  First, the linear attention of RWKV leads to significant efficiency gains but still, it may also limit the model’s performance on tasks that require recalling minutiae information over very long contexts. This is due to the funneling of information through a single vector representation over many time steps, compared with the full information maintained by the quadratic attention of standard Transformers. In other words, the model’s recurrent architecture inherently limits its ability to “look back” at previous tokens, as opposed to traditional self-attention mechanisms. While learned time decay helps prevent the loss of information,it is mechanistically limited compared to full self-attention.


emacs "gui" perfectly supports the mouse out of the box.

For use in a terminal (after using both for some time, I find the gui version better tbh), see https://www.gnu.org/software/emacs/manual/html_node/emacs/Te...


I personally think GUI mode is simply better. There is no real advantage to running emacs in a terminal imo.


Hard disagree. Sometimes the terminal is the most efficient way to run Emacs on a machine in a remote data center that has a mountain of CPUs, GPUs, memory, and disk. TRAMP can be slow and inefficient. A LSP server process on a huge monorepo can chew up a large amount of memory and CPU on startup. Running all of this stuff locally on the actual hardware your code needs to run on is great.


TRAMP is great, but to me it never feels the same as running locally. OTOH ssh -X (i.e. X forwarding) works for me for exactly this specific case (large and powerful remote dev box).

Incidentally my current setup is that I'm WFH and RDP'ing into my windows desktop, where I run VNC to our xvnc server where I ssh -X to the dev box where I run GUI emacs. It is surprisingly usable (I can also skip the RDP and run VNC directly if needed).


> to run Emacs on a machine in a remote data center that has a mountain of CPUs, GPUs, memory, and disk

But then why not run the daemon on that server and the GUI locally?


Have you actually tried that? I know the server _can_ expose a TCP socket but it's not clear to me if it's intended to work across machines.


I have been testing Emacs 28 for a few weeks now, as it was reported to improve performance on LSP integration (which was sometimes sluggish for me with rust).

I have to say I wasn't disappointed. It's now comparatively blazingly fast. I hugely recommend to give it a try if you work in emacs with LSP.

Since my distro had not packaged it yet, I installed it from guix instead (package `emacs-next`). It worked like a charm.


Do we have any idea when 28 will be available through the regular guix Emacs package?


emacs-next is how you installed lsp?

I tried installing a lsp package at some point, but I couldn't figure out if it was actually doing anything over and above the language major modes. Not sure if it was actually doing anything at all actually.


Sorry, to clarify: I used the package `emacs-next` from guix to install emacs 28. You still need to install and configure emacs-lsp along with the LSP backend(s) you want to use.

I'm currently using the doom-emacs "emacs distro" (https://github.com/hlissner/doom-emacs), so doom is taking charge of setting up emacs-lsp for me.


Thank you for that. I haven't really checked out any of these distros (doom, prelude, etc.), but maybe I should.

One of the issues I have is that I work on multiple disconnected servers, and so I usually end up installing emacs on all of them. Using tramp is an option, but often times I just want a local editor because I find myself needing one in the context I'm working in. But all the servers run slightly different OS versions, etc.

Maybe a mangaged-for-me distribution that does most everything I want without me fiddling with it is the way to go.


For me it was worked relatively well.

I track my config in a git repo, and doom-emacs is rather helpful on what external tools must be installed for various configuration "layers" (like when working with such and such language). When using LSP, it can even proposes to install an appropriate language server in some cases.


Don't think out of the box, but https://github.com/emacs-lsp/lsp-mode works rather well with emacs 28 from my experience. Emacs "distros" such as Spacemacs and Doomemacs both make using it a nearly "out of the box" experience.

(not supporting the editor/ide war troll, just trying to contribute some useful info here)


Yeah emacs was a revelation for me on this regard, that the idea of "Free Software" could be more than just providing souce code and license.

Emacs is actually explicitly designed to give you full power to explore, discover and modify it in order to tailor it to your needs and whims.


It is actually amazing, emacs is probably the best example of a "living" program you can modify interact and debug directly while you are coding it.

Obviously lisp has been designed around its REPL and debugging experience and the concept of "living programmable programs", but emacs reaches an insanely high level of business complexity, far higher than any software any of us probably writes at work and yet modifying and learning the very program you are using with the program itself is a 30+ years old solved problem that apparently ecosystems outside lisp dialects seem not to pick up.


Or that they used VC money to outcompete companies using artificially low prices, and then raised the prices to make some profit once they got a hold on the market


It's time to disrupt the disrupters.


The GPL prevents it actually (I made a sibling comment with some details)


> you'd need to prevent the people you're granting the license to from changing it in derivative works

No you don't need to. People cannot change the license of code they don't own nilly-willy, only the owner(s) of code can (which can get hairy when a codebase has multiple owners).

However a license can allow sublicensing, aka sharing the code or software under a different license, but this right must be explicitly granted by the license (or by the code owner directly), and they may be conditions attached (such as attribution)

MIT:

"[...] Permission is hereby granted,[...] to use, copy, modify, merge, publish, distribute, sublicense, [...] subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software."

GPL:

"[...] You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License."

So the MIT license grants you the right to redistribute the code under GPL (under attribution condition), but GPL does not grant you the right to redistribute the code under MIT.

(Note that all of this applies to redistributing the code or the software, not about your own internal use of the code or the software)


For emacs users, there are some modes with similar functionalities (as expected from emacs!): picture-mode and artist-mode

See for example http://xahlee.info/emacs/emacs/emacs_ascii_diagram.html


Came here to say this. If it's text, Emacs can handle it.

Oh, also, if you want your ASCII diagrams to render into pretty pictures, I found org-mode + ditaa to work quite nicely:

https://www.orgmode.org/worg/org-contrib/babel/languages/ob-...


Another option for rendering (to SVG) is goat:

https://github.com/blampe/goat


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: